accumulo-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From build...@apache.org
Subject svn commit: r907368 - in /websites/staging/accumulo/trunk/content: ./ glossary.html release_notes/1.6.0.html
Date Tue, 29 Apr 2014 18:24:04 GMT
Author: buildbot
Date: Tue Apr 29 18:24:04 2014
New Revision: 907368

Log:
Staging update by buildbot for accumulo

Modified:
    websites/staging/accumulo/trunk/content/   (props changed)
    websites/staging/accumulo/trunk/content/glossary.html
    websites/staging/accumulo/trunk/content/release_notes/1.6.0.html

Propchange: websites/staging/accumulo/trunk/content/
------------------------------------------------------------------------------
--- cms:source-revision (original)
+++ cms:source-revision Tue Apr 29 18:24:04 2014
@@ -1 +1 @@
-1587633
+1591047

Modified: websites/staging/accumulo/trunk/content/glossary.html
==============================================================================
--- websites/staging/accumulo/trunk/content/glossary.html (original)
+++ websites/staging/accumulo/trunk/content/glossary.html Tue Apr 29 18:24:04 2014
@@ -101,21 +101,21 @@
 <li><strong>iterator</strong> - a mechanism for modifying tablet-local
portions of the key/value space. Iterators are used for standard administrative tasks as well
as for custom processing.</li>
 <li><strong>iterator priority</strong> - an iterator must be configured
with a particular scope and priority.  When a tablet server enters that scope, it will instantiate
iterators in priority order starting from the smallest priority and ending with the largest,
and apply each to the data read before rewriting the data or sending the data to the user.</li>
 <li><strong>iterator scopes</strong> - the possible scopes for iterators
are where the tablet server is already reading and/or writing data: minor compaction / flush
time (<em>minc</em> scope), major compaction / file merging time (<em>majc</em>
scope), and query time (<em>scan</em> scope)</li>
-<li><strong>gc</strong> - </li>
+<li><strong>gc</strong> - process that identifies temporary files in HDFS
that are no longer needed by any process, and deletes them.</li>
 <li><strong>key</strong> - the key into the distributed sorted map which
is accumulo.  The key is subdivided into row, column, and timestamp.  The column is further
divided into  family, qualifier, and visibility.</li>
 <li><strong>locality group</strong> - a set of column families that will
be grouped together on disk.  With no locality groups configured, data is stored on disk in
row order.  If each column family were configured to be its own locality group, the data for
each column would be stored separately, in row order.  Configuring sets of columns into locality
groups is a compromise between the two approaches and will improve performance when multiple
columns are accessed in the same scan.</li>
 <li><strong>log-structured merge-tree</strong> - the sorting / flushing
/ merging scheme on which BigTable's design is based.</li>
-<li><strong>logger</strong> - </li>
+<li><strong>logger</strong> - in 1.4 and older, process that accepts updates
to tablet servers and writes them to local on-disk storage for redundancy. in 1.5 the functionality
was subsumed by the tablet server and datanode with HDFS writes.</li>
 <li><strong>major compaction</strong> - merging multiple files into a single
file.  If all of a tablet's files are merged into a single file, it is called a <em>full
major compaction</em>.</li>
-<li><strong>master</strong> - </li>
+<li><strong>master</strong> - process that detects and responds to tablet
failures, balances load across tablet servers by assigning and migrating tablets when required,
coordinates table operations, and handles tablet server logistics (startup, shutdown, recovery).</li>
 <li><strong>minor compaction</strong> - flushing data from memory to disk.
 Usually this creates a new file for a tablet, but if the memory flushed is merge-sorted in
with data from an existing file (replacing that file), it is called a <em>merging minor
compaction</em>.</li>
-<li><strong>monitor</strong> -</li>
+<li><strong>monitor</strong> - process that displays status and usage information
for all Accumulo components.</li>
 <li><strong>permissions</strong> - administrative abilities that must be
given to a user such as creating tables or users and changing permissions or configuration
parameters.</li>
 <li><strong>row</strong> - the portion of the key that is controls atomicity.
 Keys with the same row are guaranteed to remain on a single tablet hosted by a single tablet
server, therefore multiple key/value pairs can be added to or removed from a row at the same
time. The row is used for the primary sorting of the key.</li>
 <li><strong>scan</strong> - reading a range of key/value pairs.</li>
 <li><strong>tablet</strong> - a contiguous key range; the unit of work
for a tablet server.</li>
 <li><strong>tablet servers</strong> - a set of servers that hosts reads
and writes for tablets.  Each server hosts a distinct set of tablets at any given time, but
the tablets may be hosted by different servers over time.</li>
-<li><strong>timestamp</strong> - the portion of the key that controls versioning.
 Otherwise identical keys with differing timestamps are considered to be versions of a single
<em>cell</em>.  Accumulo can be configured to keep the <em>N</em>
newest versions of each <em>cell</em>.  When a deletion entry is inserted, it
deletes all earlier versions for its cell.</li>
+<li><strong>timestamp</strong> - the portion of the key that controls versioning.
Otherwise identical keys with differing timestamps are considered to be versions of a single
<em>cell</em>.  Accumulo can be configured to keep the <em>N</em>
newest versions of each <em>cell</em>.  When a deletion entry is inserted, it
deletes all earlier versions for its cell.</li>
 <li><strong>value</strong> - immutable bytes associated with a particular
key.</li>
 </ul>
   </div>

Modified: websites/staging/accumulo/trunk/content/release_notes/1.6.0.html
==============================================================================
--- websites/staging/accumulo/trunk/content/release_notes/1.6.0.html (original)
+++ websites/staging/accumulo/trunk/content/release_notes/1.6.0.html Tue Apr 29 18:24:04 2014
@@ -95,7 +95,7 @@
 <p>Apache Accumulo 1.6.0 adds some major new features and fixes many bugs.  This release
contains changes from 609 issues contributed by 36 contributors and committers.  </p>
 <p>Accumulo 1.6.0 runs on Hadoop 1, however Hadoop 2 with HA namenode is recommended
for production systems.  In addition to HA, Hadoop 2 also offers better data durability guarantees,
in the case when nodes lose power, than Hadoop 1.</p>
 <h2 id="notable-improvements">Notable Improvements</h2>
-<h3 id="multiple-namenode-support">Multiple namenode support</h3>
+<h3 id="multiple-volume-support">Multiple volume support</h3>
 <p><a href="http://research.google.com/archive/bigtable.html">BigTable's</a>
design allows for its internal metadata to automatically spread across multiple nodes.  Accumulo
has followed this design and scales very well as a result.  There is one impediment to scaling
though, and this is the HDFS namenode.  There are two problems with the namenode when it comes
to scaling.  First, the namenode stores all of its filesystem metadata in memory on a single
machine.  This introduces an upper bound on the number of files Accumulo can have.  Second,
there is an upper bound on the number of file operations per second that a single namenode
can support.  For example, a namenode can only support a few thousand delete or create file
request per second.  </p>
 <p>To overcome this bottleneck, support for multiple namenodes was added under <a
href="https://issues.apache.org/jira/browse/ACCUMULO-118" title="Multiple namenode support">ACCUMULO-118</a>.
 This change allows Accumulo to store its files across multiple namenodes.  To use this feature,
place comma separated list of namenode URIs in the new <em>instance.volumes</em>
configuration property in accumulo-site.xml.  When upgrading to 1.6.0 and multiple namenode
support is desired, modify this setting <strong>only</strong> after a successful
upgrade.</p>
 <h3 id="table-namespaces">Table namespaces</h3>
@@ -181,6 +181,7 @@ issues such as JVM garbage collection pa
 <li><a href="https://issues.apache.org/jira/browse/ACCUMULO-2519" title="FATE operation
failed across upgrade">ACCUMULO-2519</a> FATE operation failed across upgrade</li>
 </ul>
 <h2 id="known-issues">Known Issues</h2>
+<h3 id="slower-writes-than-previous-accumulo-versions">Slower writes than previous
Accumulo versions</h3>
 <p>When using Accumulo 1.6 and Hadoop 2, Accumulo will call hsync() on HDFS.
 Calling hsync improves durability by ensuring data is on disk (where other older 
 Hadoop versions might lose data in the face of power failure); however, calling
@@ -191,6 +192,14 @@ the number of concurrent writers to that
 50 concurrent writers would equate to approximately 200M of Java heap being used for
 mutation queues.</p>
 <p>For more information, see <a href="https://issues.apache.org/jira/browse/ACCUMULO-1950"
title="Reduce the number of calls to hsync">ACCUMULO-1950</a> and <a href="https://issues.apache.org/jira/browse/ACCUMULO-1905?focusedCommentId=13915208&amp;page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13915208">this
comment</a>.</p>
+<p>Another possible cause of slower writes is the change in write ahead log replication

+between 1.4 and 1.5.  Accumulo 1.4. defaulted to two loggers servers.  Accumulo 1.5 and 1.6
store 
+write ahead logs in HDFS and default to using three datanodes.  </p>
+<h3 id="batchwriter-hold-time-error">BatchWriter hold time error</h3>
+<p>If a <code>BatchWriter</code> fails with <code>MutationsRejectedException</code>
and the  message contains
+<code>"# server errors 1"</code> then it may be <a href="https://issues.apache.org/jira/browse/ACCUMULO-2388">ACCUMULO-2388</a>.
 To confirm this look in the tablet server logs 
+for <code>org.apache.accumulo.tserver.HoldTimeoutException</code> around the
time the <code>BatchWriter</code> failed.
+If this is happening often a possible work around is to set <code>general.rpc.timeout</code>
to <code>240s</code>.    </p>
 <h3 id="other-known-issues">Other known issues</h3>
 <ul>
 <li><a href="https://issues.apache.org/jira/browse/ACCUMULO-1507" title="Dynamic
Classloader still can't keep proper track of jars">ACCUMULO-1507</a> Dynamic Classloader
still can't keep proper track of jars</li>



Mime
View raw message